[a64] Implement raw clock source

Uses `CNTFRQ` and `CNTVCT` system-registers as a raw clock source.

On my ThinkPad x13s, the raw clock source returns a tick-frequency of
19,200,000 while the platform clock source(QueryPerformanceFrequency)
returns 10,000,000. Almost double the accuracy over the platform-clock!

src/xenia/base/clock.cc

@@ -21,8 +21,9 @@ DEFINE_bool(clock_no_scaling, false,
             "Guest system time is directly pulled from host.",
             "CPU");
 DEFINE_bool(clock_source_raw, false,
-            "Use the RDTSC instruction as the time source. "
-            "Host CPU must support invariant TSC.",
+            "On x64, Use the RDTSC instruction as the time source. Requires "
+            "invariant TSC. "
+            "On a64, Use the CNTVCT_EL0 register as the time source",
             "CPU");
 
 namespace xe {

src/xenia/base/clock.h

@@ -18,6 +18,8 @@
 
 #if XE_ARCH_AMD64
 #define XE_CLOCK_RAW_AVAILABLE 1
+#elif XE_ARCH_ARM64
+#define XE_CLOCK_RAW_AVAILABLE 1
 #endif
 
 DECLARE_bool(clock_no_scaling);

src/xenia/base/clock_a64.cc

@@ -0,0 +1,50 @@
+/**
+ ******************************************************************************
+ * Xenia : Xbox 360 Emulator Research Project                                 *
+ ******************************************************************************
+ * Copyright 2024 Ben Vanik. All rights reserved.                             *
+ * Released under the BSD license - see LICENSE in the root for more details. *
+ ******************************************************************************
+ */
+
+#include "xenia/base/clock.h"
+#include "xenia/base/platform.h"
+
+#if XE_ARCH_ARM64 && XE_CLOCK_RAW_AVAILABLE
+
+#include "xenia/base/logging.h"
+
+#ifdef _MSC_VER
+#include <arm64_neon.h>
+#include <intrin.h>
+#else
+#include <arm_neon.h>
+#endif
+
+// Wrap all these different cpu compiler intrinsics.
+#if XE_COMPILER_MSVC
+constexpr int32_t CNTFRQ_EL0 = ARM64_SYSREG(3, 3, 14, 0, 0);
+constexpr int32_t CNTVCT_EL0 = ARM64_SYSREG(3, 3, 14, 0, 2);
+#define xe_cpu_mrs(reg) _ReadStatusReg(reg)
+#elif XE_COMPILER_CLANG || XE_COMPILER_GNUC
+constexpr int32_t CNTFRQ_EL0 = 0b11'011'1110'0000'000;
+constexpr int32_t CNTVCT_EL0 = 0b11'011'1110'0000'010;
+
+uint64_t xe_cpu_mrs(uint32_t reg) {
+  uint64_t result;
+  __asm__ volatile("mrs \t%0," #reg : "=r"(result));
+  return result;
+}
+#else
+#error \
+    "No cpu instruction wrappers xe_cpu_mrs(CNTVCT_EL0); for current compiler implemented."
+#endif
+
+namespace xe {
+
+uint64_t Clock::host_tick_frequency_raw() { return xe_cpu_mrs(CNTFRQ_EL0); }
+uint64_t Clock::host_tick_count_raw() { return xe_cpu_mrs(CNTVCT_EL0); }
+
+}  // namespace xe
+
+#endif